Improving Object Storage Performance and Efficiency through Network Integration

现代对象存储系统是按照网络不经意的方法构建的，在这种方法中，网络被用作黑盒点对点通信介质，而存储逻辑和协议被独立地实现在客户端库和服务器代码中。不幸的是，这种方法从根本上是低效的。这种设计方法直接影响到两个核心存储操作：请求路由，将请求从客户端引导到存储节点，以及一致性复制，在节点或存储设备故障时保持数据一致性，同时防止数据丢失。新方法将设计存储感知的网络操作，可以利用存储元数据来执行网络内请求路由，并在节点和网络故障的情况下保持数据的一致性。我们将探索的创新范式有可能对技术产生广泛的影响，包括建立Web规模对象存储系统的加拿大。此外，对于数据中心（包括位于加拿大的200多个中心），该研究计划将改善基础设施的关键组成部分，并为广泛的云应用带来好处，包括生物信息学和医学信息学，在线社交网络，智能能源分配系统和文本分析等领域的应用。最后，该计划将为培训掌握尖端网络和存储技术的高素质人员提供宝贵的机会。

Modern object storage systems are built following the network-oblivious approach, in which the network isused as a black-box point-to-point communication medium, while storage logic and protocols are implementedentirely in client libraries and server codes. Unfortunately, this approach is fundamentally inefficient. Thisdesign approach directly affects two core storage operations: request routing, which directs requests fromclients to storage nodes, and consistent replication, for maintaining data consistency while preventing data losswhen nodes or storage devices fail.This project aims to explore new designs to improve the efficiency and performance of request routing andconsistent replication mechanisms. The new approach will design storage-aware network operations that canutilize storage metadata to perform in-network request routing and to maintain data consistently in spite ofnodes and network failures.The innovative paradigm we will explore has the potential to have a wide influence on technology, includingNetApp Canada that builds a Web-scale object storage system. Furthermore, for data centers (including themore than 200 centers located in Canada), this research program will improve a key component in theinfrastructure and bring benefits to a wide set of cloud applications, including applications in fields such asbioinformatics and medical informatics, online social networks, smart energy distribution systems, and textanalytics. Finally, this program will provide an invaluable opportunity for training highly qualified personnelon cutting-edge networking and storage technology.